Reliability and validity of the Dutch adaptation of the Quebec Back Pain Disability Scale.

The impact of low back pain (LBP) is strongly related to a patient's functional status.[1,2] Getting dressed, sitting, standing, walking, and cleaning the house are activities that are dependent on functional status. Physical disability and functional disability are concerns in health care, and the importance of functional status assessment is widely accepted.3 Instruments for measuring functional status arc used to measure changes in the patient's condition and to evaluate the effectiveness of therapeutic maneuvers and rehabilitation programs.2,3 According to Kirshner and Guyatt,[4] the three main purposes of measuring functional status are (1) predictive, (2) discriminative, and (3) evaluative. Predictive indexes are used to classify individuals into a set of predefined measurement categories, and, when a gold standard is available, it is possible to determine whether individuals have been classified correctly. Discriminative indexes are used to measure cross-sectional differences between individuals or groups, and evaluative indexes measure the magnitude of longitudinal differences within an individual or group over time.[4,5] In our study, prediction was not taken into consideration, because no gold standard for measuring functional status is available. Impairment measures such as spinal mobility are often poorly related to a patient's condition or level of functioning.[1,6,7] Consequently, we believe functional status should be measured directly rather than inferred from measurements of impairments such as muscle force or range of motion. Instruments have been developed for measuring the functional status of patients with LBP, for example, the Roland Disability Questionnaire (RDQ),[8,9] the Oswestry Low Back Pain Disability Questionnaire, and the Back Pain Classification Scale.[10] The choice of instruments by clinicians and researchers for measuring functional tests may be difficult, due to the variability in content examined, the difference in questions and the types of scales used, and the variable length of time needed for administration of the different instruments.2 The instruments that are available have often been developed without a well-defined conceptual model, and not all of them have been thoroughly investigated with regard to reliability, validity, and responsiveness.[10] Therefore, a new instrument for measuring functional status in patients with LBP, the Quebec Back Pain Disability Scale (QBPDS), was recently developed. This scale was designed using a conceptual approach to disability assessment and data-based methods of item development, analysis, and selection.[3] Both the French and English versions of the scale were reported to meet the most stringent criteria for a health assessment instrument.3 The English version of the QBPDS was independently translated into Dutch by two of the investigators MWvT and BWK) and a colleague with an academic degree in English. The QBPDS consists of 20 items concerning activities of daily living. Answers can be given on a six-point verbal-numerical scale, ranging from 0 (not difficult) to 5 (unable to perform). A total sum score can be calculated, ranging from 0 to 100 points.

If the Dutch version of the QBPDS was shown to have psychometric properties similar to those of the English version (ie, high levels of reliability and validity), this finding would suggest that the instrument is sufficiently robust to be used in different cultures without modification. In addition, if the Dutch version retained the qualities of the English version, this would suggest that language- and culture-specific instruments need not always be generated to examine functional status and that for practical reasons assessment of translated versions of instruments should be considered before new instruments are developed. Other successful translations of instruments into Dutch are, for example, the McGill Pain Questionnaire,[11] the MOS 36-item Short Form Health Survey (SF-36),[12] and the Dartmouth COOP Functional Health Assessment Charts/WONCA.[13] The aim of this study was to assess the reliability and validity of the Dutch adaptation of the QBPDS.

Method

Subjects

Our study was part of a larger study on the management of patients with chronic LBP in general practice by physicians, in which 650 patients participated. Inclusion criteria for patients in the study were (1) symptoms or complaints of LBP (International Classification of Primary Care [ICPC] code L86); (2) LBP with radiation, including herniation and diskopathy (ICPC code L03); (3) LBP complaints for at least 3 months at the time the study began; and (4) age between 20 and 60 years, The exclusion criterion was specific LBP caused by infection, metastasis, osteoporosis, rheumatic disorders, or fractures. The patients were selected by their general practitioners, who were all participants in the Registration Network of Family Practices of the University of Limburg in the Netherlands,[14] from their computerized databases, using the appropriate ICPC codes (L03 and L86). The general practitioners then excluded the patients who did not meet the other inclusion and exclusion criteria regarding age, duration of complaints' and cause of the complaints. The registration number, date of birth, and ICPC code of all eligible patients were sent to the Institute for Research in Extramural Medicine (Amsterdam, the Netherlands), where one of the investigators (MWvT) randomly selected the agreed number of patients for each general practitioner. If the general practitioner had selected fewer than 35 eligible patients, all patients were included.

The general practitioners sent the first questionnaire with a letter of information and introduction to the patients. We put the questionnaires into the envelopes, but the general practitioners mailed the envelopes to their patients. In this way, the privacy of the nonrespondents was guaranteed. A random sample of 210 of the 650 questionnaires included a QBPDS. We did not know which envelope was sent to which patient, and the general practitioners did not know which envelope included a QBPDS. We were, therefore, unable to trace the nonrespondents of the QBPDS study. We did, however, have information obtained from the general practitioners about the age, gender, and duration of the current complaints of the nonrespondents of the larger study, in which the nonrespondents of the QBPDS study participated.

One hundred twenty patients returned the QBPDS (response rate=57%). Table 1 shows the characteristics of the 120 respondents. The mean age of the respondents was 39.7 years (SD=10.4, range=21-60). Sixty percent of the respondents were men, and 40% of the respondents were women. The median duration of their current complaints, according to the general practitioners, was 12 months X=17.7, SD=21.5, range=0-128).

Table 1.
Patient Characteristics: Age, Number of Recurrences of Low Back
Pain, and Duration of Current Complaints'
                    %        n
Age
  X                         39.7
  SD                        10.4
  Range                     21-60
Gender
  Men              60       69
  Women            40       46

Recurrences
  0                56       60
  1                13       14
  2-5              21       23
   >5              9        10

Duration (mo)
  0-6             14        16
  7-1 2            6         7
  13-24            8         9
    >25            72       83
"Data were missing for five subjects.

The mean age of the nonrespondents (43%) was 40.3 years (SD=10.1, range=21-60). Sixty-seven percent o the nonrespondents were men, and 33% of the nonrespondents were women. The median duration of their current complaints, according to the general practitioners, was 10 months (X=25.4, SD=61.0, range=4-402).

Instruments

The QBPDS was completed by the patients at the beginning of the study (baseline), after approximately 1 week, and finally after 4 months. Patients also received the RDQ and a general questionnaire at the beginning of the study and after 4 months. The general questionnaire dealt with various aspects of LBP, such as the severity of the pain (measured on a verbal-numerical scale ranging from I to 10) and the opinion of the patient concerning the course of the LBP complaints (measured on a five-point verbal scale ranging from "much improved" to "much worse"). The RDQ is a frequently used functional status questionnaire containing 24 questions with two answer categories (yes/no). The RDQ has been reported to yield reliable and valid measurements of functional status in patients with LBP.[8,9] Roland and Morris reported a correlation coefficient of .91 for short-term repeatability. They also reported good agreement between the rating on the RIJQ and the responses on a six-point pain rating scale (agreement not quantified).8,9 Longitudinal construct validity and responsiveness, however, have not yet been examined sufficiently.

Dropouts

Dropouts were defined as subjects who returned the questionnaires at the time baseline measurements were established, but did not return the questionnaires at the 1-week or 4-month measurements. There were 20 dropouts, 10 at the 1-week measurement and 10 at the 4-month measurement. The mean age of the dropouts was 39 years (SD=11.1, range=21-60). The median duration of their complaints, according to the general practitioners, was 12 months (range=0-128). In most cases, the reasons for dropping out were unspecified, but I patient had a shoulder operation, I patient was hospitalized, and 1 patient had died.

Reliability of Outcome Measures

Test-retest reproducibility. For self-rated tests, the test-retest reproducibility is assessed by administering the scale on two occasions, separated by a time interval that is sufficiently short for us to assume that the variable being measured has not changed.[15] The time interval, however, should not be so short that patients are able to remember their first response. In this study, we used a time interval of 1 week. The test-retest reproducibility was calculated by Pearson's correlation coefficient and the intraclass correlation coefficient (ICC).[16] The ICC was calculated using a repeated-measures analysis of variance and the equation:

ICC = BMS - EMS/BMS + EMS X (n - 1)

+n[(TMS - EMS)/n]

where n=number of tests, Bms=between-subjects mean square, EMS=error mean square, and TMS=total mean square.

Internal consistency. The internal consistency of a scale relates to its homogeneity. All items on a scale designed to measure the same construct should measure different aspects of the same trait.[15] The different items, therefore, should correlate moderately with each other, and they should also correlate with the total score. Internal consistency was measured by calculating Cronbach's alpha and the item-total correlation at the beginning of the study and after 4 months. Cronbach's alpha is used to calculate the mean of all possible split-half combinations.[17] The item-total correlation reflects the strength of the relationship between a single question and the QBPDS sum score. The current opinion is that this correlation should be higher than .2.[17]

Mean-against-difference graph. To visualize the dispersion of the measurements, the mean of the two scores on the QBPDS determined at baseline and after 1 week and the difference between these scores were plotted against each other for each subject.[18]

Validity of Outcome Measures

Construct validity. Because there is no gold standard with which the results of the QBPDS could be compared, we believe that sensitivity and specificity could not be measured and other ways of establishing validity had to be used.[15] In our study, construct validity was defined as the conceptual basis for using a measurement to make an inferred interpretation,[19] and we used this approach to assess the validity of the QBPDS. Functional status can be considered as a construct that cannot be measured directly by one test item and must therefore be characterized and measured through a composite of items. The validity measurements of evaluative and discriminative indexes are not equal. We contend, therefore, that cross-sectional construct validity should be measured for discriminative indexes and that longitudinal construct validity should be measured for evaluative indexes. In cross-sectional construct validity, the scale to be examined and an external criterion should correlate at one point in time, whereas longitudinal construct validity consists of the measurement of change over time and the investigated index should correlate with an external criterion over time.[4] Longitudinal construct validity may also be considered as a strategy for evaluating responsiveness.[20]

In order to obtain coefficients for cross-sectional construct validity, the correlation between the pain severity score and the QBPDS score and the correlation between the RDQ score and the QBVDS score were calculated at the. beginning of the study and after 4 months. To determine the degree of longitudinal construct validity, the correlation coefficients were calculated between the change scores of the QBPDS and the RDQ the patient's opinion about the course of the complaint, and the change in pain severity.

Results

Instruments

Table 2 shows the mean sum scores with the standard deviations and the 95% confidence intervals for the QBPDS, the RDO, pain severity, and the course of the complaints. The minimum and maximum scores are also given.

[TABULAR DATA 2 OMITTED]

Reliability of Outcome Measures

Test-retest reproducibility. Pearson's correlation coefficient appeared to be high (r =.90). A Pearson correlation coefficient of .90 results in a multiple correlation coefficient (R 2) of .81, indicating that 81% of the variance in the 1-week scores can be explained or predicted correctly by the baseline scores. The ICC was also high (ICC=.90).

The mean-against-difference graph is presented in the Figure. The mean difference was 0.36, with a standard deviation of 8.48, which means that 95% of all the difference scores dispersed up to 16 points above or under the mean difference. If there is perfect agreement, the mean score will be zero, as the horizontal line in the graph suggests,[18] and the standard deviation will be minimal. This graph also shows a certain degree of reliability. The mean difference score is almost zero, with a moderately high standard deviation, indicating only slight differences between the baseline and 1-week scores.

Internal. consistency. Cronbach's alpha was .95 at the baseline measurement and .96 at the 4-month measurement. The high values of Cronbach's alpha, and the minimal differences between the baseline and 4-month scores, emphasize the high internal consistency of the scale. The item-total correlations varied between .58 and .83 at the beginning of the study (Tab. 3) and between .58 and .81 after 4 months.

[TABULAR DATA 3 OMITTED]

Validity of Outcome Measures

Cross-sectional construct validity. Table 4 shows the coefficients for the cross-sectional correlation. The correlation coefficients between the QBPDS and RDQ sum scores and the QBPDS and pain severity sum scores were .80 and .74 at the baseline measurement and .91 and .70 at the 4-month measurement, respectively.

Table 4.
Cross-sectional Construct Validity of the Quebec Back Pain Disability
Scale (QBPDS): Pearson's Correlation Coefficients(a) Between the Sum
Scores of the QBPDS and the Roland Disability Questionnaire (RDQ)
and the QBPDS and Pain Severity (10-Point Rating Scale) at the
Beginning of the Study (T0) and After 4 Months (T4)
                                     T0              T4
QBPDS-RDQ                            80              .91
QBPDS-pain severity                 .74              .70
(a) Two-tailed significance <.01.

Longitudinal construct validity. Table 5 shows the longitudinal construct validity coefficients measured by means of correlation between the change scores for the QBPDS and the change scores for the RDQ pain severity, and the course of the complaints. Positive scores indicate a deterioration in the patient's condition, and negative scores denote improvement. Moderately high correlation coefficients were obtained (r =.60, .53, and .35). Seventy-five percent of the subjects with change scores of 10 points (10%) or more on the QBPDS had scores that changed by more than 2 points (8%) on the RDQ.

[TABULAR DATA 5 OMITTED]

Discussion

In our study, the reliability and validity of the Dutch adaptation of the QBPDS were assessed in order to decide whether this instrument for measuring functional status can be recommended as a discriminative or evaluative index and whether the development of language-specific versions of instruments such as this is needed. Our results showed that the reliability of the scale, as indicated by test-retest reproducibility and internal consistency, proved to be high, as did the cross-sectional construct validity. The longitudinal construct validity was moderately high, probably because there was not enough dispersion in the change scores of all subjects on every scale.

The data in Table 5 suggest that, on average, there was little change in the participating patients. This finding is misleading, because improvement and relapse cancel each other out and result in average change scores around zero. The fact that the longitudinal construct validity is only moderate does not necessarily imply that the scale is unable to detect change over time. Our study population consisted of patients with chronic LBP, a population that appeared to be suitable for determining the reliability and cross-sectional construct validity of the QBPDS. The longitudinal construct validity of the questionnaire, however, was difficult to evaluate, because large changes in functional status are not likely to occur in this type of population.

The functional status of the patients in our study changed (ie, the QBPDS scores of 95% of all participants changed between 0 and 24 points), and these changes were considerable. The status of many people changed in a similar way, so no high correlation coefficients could be expected. That is, although the change in the condition of the patients seemed to be considerable, there was not enough dispersion in the change scores. Seventy-five percent of the subjects with change scores on the QBPDS of 10 points (10%) or more had changes of more than 2 points (8%) on the RDQ indicating that the QBPDS might be more responsive than the RDQ. We conclude that the QBPDS seems to be a suitable discriminative instrument, but its usefulness as an evaluative instrument could not be adequately assessed in our study.

Apart from the need for reliability and validity, several other features of measurement instruments are important, such as applicability, practicality, and comprehensiveness.[17] Applicability depends on the appropriateness of the content and the emphasis of the measurement with regard to the purpose. Practicality refers to the burden on the respondents and the professionals, the costs, and the method of scoring. Comprehensiveness means the completeness of the measurement and the extent to which the topic is covered. Kopec et al[3] demonstrated, by methods of content development, extensive psychometric evaluation, and item selection, that the QBPDS meets the criteria of applicability and comprehensiveness. We agree with Kopec et al[3] that the QBPDS is a simple instrument that is easily self-administered, and that the costs are low. Despite explicit instructions to complete all 20 items, however, 10.8% of the patients had one or more missing values at the baseline measurement. Similarly, 9.2% of the patients did not complete the RDQ We therefore argue that this flaw in practicality is not a specific characteristic of the QBPDS, but is inherent to all self-administered, mailed questionnaires. In practice, where questionnaires can be filled in during a consultation, this flaw may be a minor problem, because completeness can be checked immediately.

Kopec et al[3] stated that the English and French versions of the QBPDS were highly reliable, valid, and responsive in their Canadian population of ambulatory patients with back pain. Kopec et al reported an ICC of .92, whereas in our study the ICC was .90; Cronbach's alpha coefficients were .96 and .95/.96, respectively. The cross-sectional construct validity in the Canadian study and in our Dutch study appeared to be high. The correlation coefficient between the QBPDS and the RDQ was .77 in the Canadian study and.80 at the baseline measurement and .91 at the 4-month measurement in our study. Therefore, the results of our study support the findings of Kopec et al to a large extent, which provides stronger evidence for the usefulness of the QBPDS.

Our study shows that it is possible to translate a functional status questionnaire into another language, in our case into Dutch, without losing the psychometric properties of the original (English) version. Thus, translating existing scales appears to be feasible, and is clearly much more efficient than developing a new scale. Translation into different languages, and subsequent validation of questionnaires, is of importance for international understanding of the measurement properties of these scales when they are used in different cultural settings. The positive experiences with the translation and validation of the well-known SF-36 in numerous different languages support this idea. Finally, the use of the same questionnaire in different countries will enhance the comparability of the results of, for example, randomized clinical trials published in. the international literature.

We agree with Kopec et al[3] that the QBPDS is a reliable and valid instrument for assessing functional status in patients with LBP. Our study showed that the QBPDS is a suitable index for discriminative purposes (ie, discriminating between individuals or groups at one point in time). Whether this instrument is also suitable for evaluative purposes in practice (eg, for the assessment of the progression of a patient after treatment) could not be assessed in our study. Kopec et al[3] showed promising results of the responsiveness of the QBPDS.

Conclusion

Because the QBPDS seems to be a valid and reliable questionnaire for the assessment of the functional status of patients with LBP, we recommend the use of the QBPDS in future clinical trials and for comparing different individuals or groups of patients with LBP. The efficiency of the QBPDS in evaluating longitudinal change in an individual or a group (ie, the evaluation of the progress of patients in practice) seems to be promising and, we believe, should be the subject of further research.

[Figure ILLUSTRATION OMITTED]

References

[1] Deyo RA. Measuring the functional status of patients with low back pain. Arch Phys Med Rehabil 1988;69:1044-1053. [2] Bombardier C, Tugwell P. Methodological considerations in functional assessment. J Rheumatol. 1987;14(suppl 15):6-10. [3] Kopec JA, Esdail JM, Abrahamowicz M, et al. The Quebec Back Pain Disability Scale: measurement properties. Spine. 1995;20:341-352. [4] Kirshner B, Guyatt G. A mediodological framework for assessing health indices. J Chronic Dis. 1985;38:27-36. [5] Guyatt GH, Kirshner B, Jaeschke R. Measuring health status: What are the necessary measurement properties? J Clin Epidemiol. 1992;45: 1341-1345. [6] Flor H, Turk DC. Etiological theories and treatment for chronic back pain, : somatic models and treatment. Pain. 1984;19:105-121. [7] Millard RW. A critical review of questionnaires for assessing pain-related disability. J Occup Rehabil 1991;1:289-302. [8] Roland M, Morris R. A study of the natural history of low back pain, part II. Spine. 1983;8:145-150. [9] Roland M, Morris R. A study of the natural history of low back pain: part I. Spine. 1983;8:141-144. [10] McDowell I, Newell C. Measuring Health: A Guide to Rating Scales and Questionnaires. New York, NY: Oxford University Press Inc; 1987:12-35. [11] Vertommen H. De McGill pain questionnaire anno 1988. In: van der Kloot WA, Vertommen H, eds. De MPQ-DLV: Een Standaard Nederlandstalige Versie van de McGill Questionnaire. Lisse, the Netherlands: Swets and Zertlinger; 1989. [12] van der Zee K, Sanderman R, Heyink J. De psychometische kwalsteiten van de MOS 36-item Short Form Health Survey (SF-36) in een Nederlandse populatie. T Soc Gerondheidsz. 1993;71:183-191. [13] Scholten JHG, Van Weel C. Functional Status Assessment: The Dartmouth COOP Functional Health Assessment Charts/WONCA. Lelystad, the Netherlands: Meditehst; 1992. [14] Metsemakers JFM, Hoppener P, Knottnerus JA, et al. Computerized health information in the Netherlands: a registration network of family practices. Br J Gen Pract. 1992;42:102-106. [15] Streiner DL, Norman GR. Health Measurement Scales: A Practical Guide to Their development and Use. New York, NY: Oxford University Press Inc; 1992. [16] Fleiss JL. The Design and Analysis of Clinical Experiments. New York, NY: John Wiley & Sons Inc; 1986:1-32. [17] Deyo RA. Measuring functional outcomes in therapeutic trials for chronic disease. Control Clin Trials. 1984;5:223-240. [18] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. February 1986: 307-310. [19] Task Force on Standards for Measurement in Physical Therapy. Standards for tests and measurements in physical therapy practice. Phys Ther. 1991;71:589-622. [20] Deyo RA, Centor RM. Assessing the responsiveness of functional scales to clinical change: an analogy to diagnostic test performance. J Chronic Dis. 1986;39:897-906.

EM Schoppink, MScPT, is Research Fellow, Institute for Research in Extramural Medicine, Vrije Universiteit, Amsterdam, the Netherlands.

MW van Tulder, MSc, is Research Fellow, Institute for Research in Extramural Medicine, Vrije Universiteit.

BW Koes, PhD, is Senior Investigator, Institute for Research in Extramural Medicine, Vrije Universiteit, van der Boechorststraat 7, 1081 BT Amsterdam, the Netherlands ([e-mail]). Address all correspondence to Dr Koes.

AJHM Beurskens, MScPT, is Research Fellow, Department of Epidemiology, University of Limburg, Maastricht, the Netherlands.

RA de Bie, MScPT, is Research Fellow, Department of Epidemiology, University of Limburg.

This study was supported by a grant from the Dutch Health Insurance Executive Board.

This article was submitted January 10, 1995, and was accepted October 30, 1995.